Looking for a job

Shubham Jakhmola

@shubhamjakhmola

Message

I build production agentic RAG systems, evaluation pipelines, and reliable AI features.

India

Message

What I'm looking for

I'm looking to build production AI features where I can own agentic workflows, retrieval, evaluation, and monitoring from prototype through real-user deployment.

At Bobble AI, I replaced a rule-based chatbot with a LangGraph agent using agentic RAG on Azure for a consumer keyboard app serving 5M+ daily users. The feature became the default experience, raising onboarding completion from 20% to 60% and chat-feature DAU by 150% QoQ.

I build the retrieval, evaluation, and monitoring foundations that make AI features trustworthy in production, from hybrid search and LLM-as-judge evaluation to Prometheus/Grafana monitoring. Earlier, I built PySpark and Databricks pipelines processing 20M+ records daily and delivered demand forecasting that reduced stock-outs by 18%.

Experience

Work history, roles, and key accomplishments

AI Engineer

Bobble AI

Jun 2024 - Aug 2025 (1 year 2 months)

Rebuilt a rule-based chatbot as a LangGraph tool-calling agent on Azure with agentic RAG, improving onboarding completion from 20% to 60% and chat-feature DAU by 150% QoQ. Built evaluation and monitoring systems that reduced thumbs-down rate by roughly half.

LangGraph Azure RAG Hybrid Search Reciprocal Rank Fusion Redis Ragas LLM As Judge Prometheus Grafana Python fastAPI Agile

AI Model Trainer

Translated

Sep 2023 - Jun 2024 (9 months)

Crafted prompt-answer pairs and preference rankings for RLHF/SFT on LLMs, improving alignment, safety, and multilingual fluency through rubric-based reviews and edge-case rewrites.

RLHF SFT Prompt Engineering LLM Evaluation

Data Engineer

Northumbria University

Nov 2022 - Aug 2023 (9 months)

Automated ETL and data-quality checks for 30k+ student records using PostgreSQL and Power Query, reducing monthly reconciliation time from two days to under three hours.

ETL PostgreSQL Power Query Data Quality

Research Assistant

Newcastle University

Aug 2022 - Nov 2022 (3 months)

Built a semi-automated labelling workflow for ~10k tree crowns across 50 km² of aerial imagery, improving a U-Net segmentation model from 0.65 to 0.83 F1.

ECognition QGIS U Net Image Segmentation

Python Developer

Golden Pegasus IT

Aug 2019 - Aug 2021 (2 years)

Engineered a medallion-style ERP data pipeline in Databricks with PySpark and Delta Lake, ingesting 20M+ records/day and cutting latency from 6 hours to under 45 minutes. Developed demand-forecasting models that reduced stock-outs by 18%.

Python PySpark Databricks Delta Lake ETL Demand Forecasting Machine Learning